Transcript
Can you confirm?
I show live, I see us, so I hope it's true.
I don't see anything that says we're live.
Oh, there it just changed.
Yes, I see us, so people should see it now, people in the chat.
If you're here in a chat, you should us.
I can see some of you doing things.
Just, hey, can hear you great, good, wonderful, we're live.
We're not starting yet.
we will start at the top of the hour.
When we do live video events, We always get on early just because there's a lot of opportunity for things to go wrong.
And we don't want to waste your time.
So we get early and make sure it works.
Always, always a good idea.
I'm trying to figure out a pop the chat window out.
It used to pop out separately.
Yeah, you can.
There's like a it's a little menu though.
I know we've got some people here already, which is great.
And we'll probably have to repeat this when we start.
But just since you guys are here, I'll mention it now.
The whole point of this is to take stories from you, guys, and talk about how we're going to split them with AI, but also get Mike's insight.
We are using YouTube's question system, it was new to me, so it might be new you.
It is in the chat.
You can see a little question button.
I'm hoping you can submit your stories there.
That means it will make it easier for me to queue them up.
Instead of just putting them into the chat, there were some that were in the chapter we started before we start it.
I grabbed those, so we should be good there.
But if you don't mind trying to use that question queue thing, that would be awesome.
So that'd be a big help for us.
And yeah.
Oh, and another common question is, will this video be replayable?
Our intention is to publish it on the channel, just like the rest of our YouTube stuff.
We'll put it up after we're done.
Or I think it'll just goes up.
I think it happens, right, from what you were saying earlier?
If you do something to prevent it, I it goes up.
We haven't done a lot of the live events on YouTube.
we've done alot of live webinars, but I last one was probably a year ago that we did this.
It has been a while.
I see there's a chat comment saying that they can't hear.
i think everyone else can hear us, so that might be just a turn up, unmute or turn off the volume kind of situation.
Yeah, Shay just hit your volume, probably.
Katie, you're early.
I'm always early, I was raised, if you are not early or late, always my dad.
So I, of course, passed that curse on to my daughters, too.
so I try to be early for everything.
And yes, you can claim SCUs in the Project Management Institute.
So I don't remember what categories things go in there these days, but obviously.
Katie, can you submit more than one?
Oh, submit only this story or the story with the acceptance criteria?
Great question.
You can do both.
The tool that we're using today, we do have room for you to put in your acceptance Uh, that's great.
Please do.
If it's not super obvious, what part is which?
I mean, usually I can tell, but I do split them up.
So do please let me know if it sounds super-obvious.
Um, and there's a question about how detailed, uh, it can be super high level.
I'm mean part of this is like a test, right?
We're playing with some new tech here.
We want to see how well it does.
Just thought about something.
YouTube has a pretty severe limitation on length, right?
So there's a character limit for sure.
So we may be playing with that a bit.
Yes.
We may have an issue with app next time we do this.
Maybe we have, um, and you suggested Hunter, maybe we had our own queue that people can go to because of the limitations.
I'm just looking to see if I can adjust that here.
I would doubt YouTube lets you adjust that.
They're not big on the customer.
You can adjust the sort of jail that you go into if you type too fast, but that's about it.
Oh, for if, you submit too many questions in a row.
So yeah, exactly that they let you, they, let, just the timeout.
so you can say this person is not done for X seconds, that is customization.
That must be a nasty problem.
Shane, I'm glad you can hear us.
Yeah.
Jurassic, yes, we do.
This should just go.
We're planning to stick it on the YouTube channel.
So we've got a lot of other videos up there.
Mostly are the sort of traditional produced YouTube video type stuff that Mike does.
But we're going to hook this up here too.
Can you repeat what tool you'll be using?
Yeah, guess we'll talk about this at the top also.
Happy to talk around for a second.
so um i would consider it wait wait we didn't build it i built it kind of i mean you built you builded out of there's a lot of pieces here that were not
built by me but um but it's the tool that we have put together it I would considered a prototype and part of us doing this here today is to get some real
world feedback and a spoiler alert.
Our intention is, to share a link at the end of this video with you folks to, get access to it yourselves for you to try it.
So you will be able to do it yourself at end the video.
If you choose to and that's free, we're not selling anything.
We're just hoping people try this out.
Maybe like two thirds are like really, maybe like 50% are really good.
20% or so are ones that I wouldn't have thought of, which is nice.
And the other 30% were like, whoa, what was he thinking, right?
You know, it's like obvious one or that's not needed one.
So, but it is helping.
It is hoping fine things.
Jennifer, I see your question there, your story there.
If you wouldn't mind, if you could submit that as a Q and A.
I know that folks, especially if, you're not on YouTube, it's super, super all the time.
So there's a question queue actually that we're using for this.
You should see it at the top in a little black bar where it says Q&A.
That will actually let you submit it into our queue.
It makes it just easier for us to manage.
Um, and yeah, I realized it's not a super common thing.
So, you know, You wouldn't necessarily need to do it.
Sorry about that.
We're, we're trying to manage it that way.
And Jen noticed you did tell her that good morning, Shirley and a gamer's place.
Um I think after I saw it, yeah.
I'm your dad's early too.
and you went to the wedding when they're setting up the chairs.
Yeah, haven't been that bad.
My worst was I was going to be about a minute late, probably a seriously a minutes late for a doctor appointment.
because I was stuck behind train tracks and I'm sitting in my car at the train track, looking over at my phone, wanting to call my doctor to say I would
be a minute late.
And this is one of those doctors that's always like 30 minutes late himself.
It really took all my self-control to not call the doctor and say, I am going to be minute.
I just pictured the office staff laughing at me.
So that is what kept me in check.
Scott, yes, it'll just not invest.
Oh, no, did we add that in 100 at the end?
Yeah, so well, I guess we're going to start in a minute and we'll talk in specifics.
But it's using so the spider technique, which is from Mike's Better User Stories course, but then it also instructed to use the invest method when when
writing stories.
So I'm trying to cram all that on there.
Yep.
Not so sunny Scotland.
Yeah, Tracy, me too.
I'm looking at rain outside my window today.
Had about four days of really nice, well, not really, nice but sunny weather where I could at least not wear a jacket inside the house and got teased because
now it's back to crappy weather.
So.
We're like seconds away here.
Yeah, ethical concerns regarding AI.
Clearly a big topic these days and I can only imagine will become a bigger one.
Something that we all need to keep in mind as we figure out how to use these tools, how they fit in with what we do.
Yeah.
Good point.
Are we at the top of the hour?
We are.
We should just get started.
Let's do it, let's go ahead and get started.
First off, I wanna welcome everybody here.
We are going to try splitting some stories with a tool that Hunter has been working on.
Hunter Hillegas is our Mountain Goat CTO, and he's been experimenting with artificial intelligence to write stories.
we've been using it ourselves for a little bit.
certified scrum product owner courses for them to help use this.
I saw Dave's comment here about AI replacing labor and I think where I see an AI helping in story splitting is trying to cut down on the amount of time
that people spend in things like backlog refinement.
Being able to split a story is not, I don't think it's a science where you can just look at a store and go, boom, there are the five things.
Here's the algorithm.
It's always these five thing.
So there's little bit of art to splitting the story, but I can make art, to some extent.
I think you could help with that.
Never say never, never say ever.
But I don't think I'd ever want to trust it 100%. Like, hey, just split the stories and I'll look at them.
It just dumps them into JIRA and the team works on them, I know that we'll ever get to that point.
You know, hard to say things 20 years into the future.
I think that I always want a pair of human eyes on stuff, especially now.
And we're going to try some here in a minute or two.
As I've done them, I have been impressed with a lot of what it comes up with.
It surprised me with the few splits that I wouldn't have thought of.
I might have though about them down the road, but I would have not thought about it in an initial splitting session.
All right, make the splits right now.
So I'm impressed by that, and I was surprised by the occasional hallucination.
Well, that's a weird one, right?
Where did it come up with that?
One that I saw recently was where we asked it for something like a member doing something, and all of a sudden it came up like the admin of the site doing
the thing.
And I was like, well, it's kind of weird.
So we'll probably see a few weird ones.
I hope we do see some weird as we split some stories today.
so I want to thank everybody.
and I'm going to turn it over to Hunter.
Do you want talk about the tool a little bit?
Let's talk about a few things.
So I think, first, I'll just start off with a little bit of info about how you can submit your stories.
Maybe you could be thinking about that while I talk a bit more about the tool.
There is a question, though, Mike, that is in the chat that might be worth briefly addressing is, why do you split stories?
Just a quick reminder of why is this good?
Yeah, Why split Stories?
The reason for me to split stories is so that we can gauge our progress.
There is a whole joke in the software industry called the 90% syndrome.
And it has to do with thinking we're 90%, right?
Hunter, how done are you with our story splitter tool here?
92%. 92%. I'm going to ask Hunter that question in a week, and he's going say 92% again.
And the issue there is that he is making progress at the rate at which he discovering the problem is bigger, right?
And so what we want to do is we wanna split things into really small pieces.
So if I am not asking Hunter a really stupid question, like how done are you with story splitting, but I can ask him a more reasonable question like,
hey, Hunter, how down are with training the story splitter at using the invest acronym?
at splitting stories hunter have done to you with that.
Well, it's in there.
We'll see how these tests go.
Well then, I would say it is 100%, right?
Maybe I should have said initial training, right.
But you've trained it on that.
Maybe what we're going to learn is we need to do more.
So we split stories so that we can put stories into one of two buckets.
Done, not started, we don't want to be telling people we are 92% done.
When we break things down small, We can finish those with iteration and they can quickly go from not-started to done, that's the goal in splitting.
Perfect.
Okay, let's talk a little bit about how you can submit the stories.
I know some of you already have, which is great.
You should see in the chat, we're using a YouTube feature that was new to me, but they do allow us to have a question queue.
Up at the top in chat you should still a black bar that I'm looking at.
It says Q&A, submit your story, and we'll try to split it.
Now, if you click on that, you see a button where you then submit it and it goes into this special queue, it just keeps them separate from the chart,
making it easier for us.
We do appreciate if Let's see a couple of questions and some information about how it works.
So it's very much a prototype in the sense that, you know, one of the reasons that we're doing this event and that will be sharing it with you to try for
yourself is to get more feedback at the bottom of each set of splits is a feedback form and And, you know, You can use that to give us feedback about how
well it did.
In my own experience I found some that are quite good, some, that aren't as good and places where it can definitely improve.
One of the things that I added a little just about a week ago was an explanation for each split.
So I wanted to know why it chose what it did.
And we've added that in, which actually helps a bit with understanding why its making the choices that it's making, because sometimes this can be somewhat opaque.
I think we'll probably look at those a lit bit today too, just to see what its thinking.
Im going to use the word thinking, I know its not really thinking so forgive me.
That's just the term that comes to me most easily.
So somebody asked in the chat if we were using an existing model.
Yes, we did not build our own model from scratch.
This is using models from OpenAI.
currently with a combination of some reinforcement learning, a lot of really crafted prompting, and a bunch of different finely tuned examples.
So a Bunch of Different pieces going going into that but did use an existing public model tried to experiment with quite a few of their different models
and still are really because there's a Lot of options if you've looked at theirs or models from other vendors, A lot Of different options between models
and some of those settings.
So a work in progress.
Let's see.
Anything else that we need to share?
I think before we start, I did want to mention one other thing, which is another live event that were having.
I'm going to put it up on the screen here.
Mike's doing a Better User Stories webinar, actually one week from today.
If you want sign up for that, the URL is there on on page on that slide there.
And it's free.
I don't know, Mike, if you want to give the 10-second explanation of what the bus webinar is.
We're just going to go over three things that I think can really help a team get a whole lot better at storage.
One is splitting.
So we'll talk about techniques for splitting, but we've got two other techniques in there.
And we will go with that, do an hour of webinar and hour or so of Q&A.
Join us for that if we can.
As all Chronicles, I'm glad you like our name.
we're not always good at naming things.
So someone's asking about which model.
Right now, what we're going to see today is based on OpenAI's 03 Mini.
That may not be, as this tool evolves, that may change.
So honestly, the way that we think about this is that you shouldn't have to care about what model we are using behind the scenes.
But for the stuff that were seeing today, it's the 03 mini model from Open AI.
Oh, OK.
So with that, let's get started.
I'm going to put this up on the screen here.
And we've got some that I loaded up from the chat as we'd been going.
There were some in there, and there's more to add.
We'll just keep these rolling.
When we submit a story, it does take a second or two to come up with all the different splits, so I preloaded some of these just to save us that time.
For this one, I'll read it off, then we can talk about what it came up.
The story submitted was as a user of a secondhand advertisement website, for example, Craigslist.
I want to create a publishing listing so that other people can see and respond to my advertisement.
And the way that this is formatted is breaking down stories based on the spider technique.
So spiders spike.
path, interface, data and rules.
So we'll go through these and we will talk about them.
Let's start off with the spike.
It's determined that no spike is needed and it says the story has a clear and well-understood flow with minimal excess uncertainty.
What do you think about that, Mike?
First off, there's a question, is there another way to submit a story from Laura?
Laura, the Q&A is the main way.
I don't know if we have any other way do that, but the q&a should work.
And by the way, a comment, I have not seen any of these results.
So it's not like we've queued this up and I've got You know, we picked good stories or anything like that.
I guess we tried to pick stories that we think will split well, but these I think are all going to come from you.
So I have not seen these answers.
These are new to me as we get these.
Um, I like the path split here.
This is, um, this is good.
the same way with the breaking out the title description, price and images.
And what I like about that is if we put that story itself into the system, I don't know that we need to, but if.
We put.
That in.
I would hope that.
we would get separate stories that would be things a story about title, a.
Story about description of story, about price.
and store about images, especially the one on images I'd be interested in is like breaking.
Out as a separate story.
The other three or four there might not need be broken out, But that's how we'd break that out.
I might have had that as a story where I can just enter the, you know, details about an ad and then have acceptance criteria of those things.
But I think either way works on here.
So I'm happy with that.
I.
Think the published story is good.
We probably have that published.
Story have some stories or acceptance.
Criteria about setting dates on there for one of you.
Can I set the pub date in the future?
How far into the.
Future might be something that I would that The interface ones, we're trying to get the idea of interface splitting better in Lumberjack.
The Interface splits often seem to devolve into giving iOS, Android, and a web version.
And there's more to interface-splitting than that, so we've been sending it some some more training on that.
I'm not sure that I would have an interface split here.
If I did, it would be dependent if the user interface were complicated and have, you know, do a simple version of the UI all the way down and then do the,
kind of, polish the UIs a separate story.
That's all I could do there.
Data, the...
Excuse me one second, I got to clear my throat.
No problem.
One of the things I'll say while Mike's doing that is that one of things with the interface that was something that we're continuing to tune is it really,
really wants to do a web iOS Android split when it comes to the Interface.
Like it's really hard to persuade it not to only do those.
It's like great job done next.
So it is something I would like to see improve in that area.
I think the world is so trained that interface means the user interface.
It's like, of course, split it that way.
Yeah, it's hard.
The data one, I thinks these are very similar to paths.
This is where one of the things that when I was creating the spider acronym for splitting stories, that's what we're seeing here,
the spikes, paths, interface, data, and rules.
I never meant them to be mutually exclusive.
So I would look at a list like this and go, oh, like the data ones better than the path ones or something.
I would choose the data ones.
For example, I'm not thinking about which I'd prefer here.
But I wouldn't expect this to be mutually exclusive.
So we like the Data ones here better than the Path ones?
I picked the Date ones, put those in there.
And I guess I probably do like to Data.
I do the Datasplit ones her.
Rules, i'm OK with it not finding anything on rules, but I might have wanted to do something about I'm required to enter a photo or I am required put in
a publication date for it or the publication is always today unless otherwise specified.
So I might have added a few rule stories there.
But I wouldn't expect a lot of people to have done that either.
Just a quick little note.
So you'll see underneath any of these splits, there's this little explain disclosure triangle.
You can click on that and that's an example of it explaining how it did, why it, did what it didn't.
And so it's sort of in an effort to understand the why of some of this, you can see it identified two sets of baby complexity and work that way.
Those are available anytime you see one of the splits.
OK, moving along.
Let's go to the next one.
All right.
And I'm sorry for some of these that were in the chat.
I don't have a log of who submitted them, so we appreciate you submitting them.
As a user, I want to log in using OAuth so I can access my account securely.
Then I'll just scroll this down here.
What do you think about some these, Mike?
Logging through OAauth so can set my accounts securely, paths, I was muting my mic for a second there, so I missed the first part.
OK.
As user, I want to initiate OAuth login by selecting my preferred route.
I like that as a path.
How I'm logging in.
Yeah, those are pretty obvious ones.
So I don't think we gain much there.
But it would be nice to have an AI write that for me instead of have to do that myself.
For acceptance criteria, I'd want the first one to say which providers we're supporting.
Is there a limit to those?
The interface ones, again, are pretty boring here.
These are legit.
I might have to do these things on all of those.
So I think they're legit stories.
And I don't know that I have anything other than that.
Right.
Not like AI saved me any time by saying, oh, don't forget to have an Android app.
Data.
I don' know that those stories really give us anything.
Do you think they do, Hunter?
I mean, they kind of are almost like restated versions of some of the path stories.
Exactly.
But they're pretty close.
They're not that distinct.
Yeah, I like the path ones better in this case, whereas on that previous story I liked the data ones, better, which actually I would consider that good
news that one approach isn't winning all the time.
And the rules one is nice about the clear error messages.
I don't know the first one has seen.
OK, I think we could look down this list and have the right set of stories from here.
So it's not a story that needs a whole bunch of splits.
It's a quick note on a couple of chat comments.
There was a question about access to this.
We do plan to give you guys here access to this to play with as the prototype after the event.
So there'll be a link that you can use.
There's a question about and that that's free.
It will be used be able to use free a charge the Prototype.
I mean, we're not exactly sure where this is all leading but The thing that you'll get today, you can play with for free.
Securely part, it is running on our infrastructure.
So there is a note here saying like, please don't give us your top secret information.
You should just be aware of that.
We're not in there digging around with your stuff necessarily, but just keep that in mind if that's a concern to you.
We understand that for some of these kind of tools, people do want to be able to use them inside their own infrastructure or whatever.
But that's not something we have an answer for today.
So just keep that in mind.
Yeah, I would say we're ethical and lazy, right?
We do have good ethics and we are a little too lazy to want go in there and say, oh, let's see if somebody submitted a billion dollar business idea we
can steal.
We're not going to go and look at things like that.
Here's the next one, and this was from Issam who submitted it.
As a campaign manager, excuse me, I want to configure and apply different discount percentages and periods based on the sales channel and system so that
we can tail promotions more effectively.
So I'm going to scroll that one down, Mike, if that's OK.
I like that I was I I.
Was nervous at first when campaign Manager and my mind went to politics.
I'm glad we're not there.
So spike, investigate and document the discount configuration requirements across different sales channels.
Yeah, this would be a good spike to research what type of discounts and such we need.
I think that's a Good one.
Path, I can select a sales channel and configure as discount percentage, active period.
That one's OK.
Can you score back up, Hunter, to what we put in there, this story?
Percentages and periods.
OK, so yeah, the path one is good.
What I was wondering about there was, was it broad enough in that are we trying to capture discounts?
Are we try to to by one, get one free things?
And the story didn't really say it's about buy one get discount.
So it was more about percentage discounts.
is a good one on a percentage basis.
I think it'd be interesting if we tried it sometime with a story, kind of a higher level story about, as a campaign manager,
I want to offer a variety of discounts to my customers, such as buy one, get one.
X percent off, things like that.
So that might be an interesting trial for a high level of story.
And that would probably give us more meat in the interface part, or in that path part.
So I think those are good interface.
Ooh, it's not iOS.
So let's see what these are campaign is using sales channel interface, oops, sorry, guys, I clicked on the wrong thing trying to lead it up and I screwed up.
Sorry, here we're back.
Using the sales chat on your face want to configure unique discount percentages.
I don't know if this is good or not, because I the screen where I configured sales channels.
It says, well, alternative splits by a platform were considered the story context focused on sales, channels, and systems rather than device specific interfaces.
Thus, the splits focus on tailoring discount configurations through the sales channel interface and through this system integration interface.
Yeah, I mean, this might be one where adding some more acceptance criteria information in there might have given a better output.
Yeah, I would look at this, and I'm trying to put myself in the position of the person who wrote the story initially each time.
I might look this and go, OK, that story is not good on its own, but it gives me ideas.
Because I think what it's saying there is go in, pick the interface, picked the channel you want to do, then pick discounts for the channels.
So it kind of saying associate those.
It's OK and give me an idea, But I doubt I'd copy and paste that one into Jira.
Data, configure discount with a specific percentage and period.
Yeah, those are good.
I think the data ones are.
Good data have been good consistently so far.
Campaign manager, configured discount percentages so that each promotions.
What do you all think in the chat?
Is it doing okay so Yeah, there's just some comments on presentation.
Yeah.
It would be nice if we could keep the story up while we were going through.
That's good to know if do one of these again, for sure.
So bear with us there, I guess.
Interface switch for us next time.
Max, this needs the invest criteria.
Are some of this where it didn't follow invest?
That would be good feedback, too, if you end up trying this out on your own.
It's because it's supposed to, right?
I mean, basically, the way that those rules are set up is it supposed try, but it is allowed to break those roles if it thinks it needs to.
But is that always right, unclear.
So this is an area we're getting more feedback.
What you're seeing is the tool.
So what you are seeing here is what will you be able to access after this session.
And there was a question about how long will we have access to unlimited trial expiration.
Again, this is a prototype.
It's not really a product.
We're putting it out there to get some feedback.
What we are going to give you doesn't cost you anything.
There are no limits other than some like basic rate limits, you know, some sanity, but there's not like you can do 10 stories and that's it or anything
like that.
We want you to use it and provide that feedback.
So I hope that answers that question.
As far as what we're gonna do with it long-term, We'll see.
There's a question in here.
Does this tool look at how to use all eight ways to vertically slice stories?
I'm willing to say interface data and rules.
So this is focused on Spider, which if you're not familiar with, Mike, I don't know if we talked about that enough already,
but it's the Spider technique that you came up with.
Yeah, Patty, when I was working on something for one of my clients.
It was, I don't know, 10 years ago.
A comment for people to write an article that said 12 ways to split stories.
And the next week, you'd say 15 ways right stories and 20 ways write stories, and I thought about it.
I said, well, don t have 20 Ways to Split Stories.
Even if I did have twenty ways split Stories, wouldn't remember them.
So I took over a thousand stories like 1100 stories, I printed them out, cut them on a little piece of paper, taped them up all over a wall I had.
I have this huge wall in my office and taped him up in groups trying to sort them into the smallest number of ways to split that I could come across.
And since then, i have not found stories that i couldn't split with one of those five techniques.
So I was not trying ways to vertically slice stories.
So I don't know if you're saying that we're missing eight techniques or we missing three because we only have five or there's some overlap or missing six.
But if want to send us information on what you were thinking about, we can look at training this on something further.
I was always able to split a story well using the five techniques and I didn't want 12 or 13 is too hard to remember.
So, yeah.
And incredibly, yes, it's not starting the conversation.
As far as I see some other notes, it's hard to read these as they go by.
I know, I'm sorry.
This is the downside of trying to demo a tool with a lot of text when there's a bunch of people reading at different speeds and whatnot.
My recommendation would be give it a try once we're done.
You can peruse up your letter.
We're going to try to move it along here and let's try and get to another one.
The story here is as a user, i want to search for items in a database and gets relevant results.
Okay, I would critique the story a little bit first because I wouldn't really want to include the word database in the store.
I was trying to think about my mom and my Mom was completely non techie.
the database part out.
Spike, I agree that nothing necessary.
Path, enter in a search query field, view the results.
be OK with those, but those are pretty boring.
There's probably a lot more that we would want to do there.
I'd probably want I have that I can do a basic search.
Maybe that's going to come down below.
So we'll see.
Probably have something I do.
A basic Search where I just type something in and something else that.
Can sort the results, I filter the result, so I want see things where perhaps I was searching for some new slacks yesterday,
and I could go on there and pick that I wanted hiking slaks or casual wear.
I want to narrow it down by category, something like that.
So I'd probably want see some more options there, depending on what we were doing.
And we don't know any of that because the story is so short.
But with more context, I think it might have done better with those.
interface.
Just with the other one same responsive interface?
I don't recall that on the Other Stories.
So, no, the Yeah, now I It's weird that this one's on a responsive UI and the rest aren't.
So it's got our normal split here.
I think that's okay, but this is where I would probably have a simple and a complex.
You want to search for it, I'm looking for slacks and I go in and type hiking pants on the website versus I going to type pants and i can filter it by
active wear, hiking, evening wear or something like that, whatever.
So I do think there'd be more we could do with interface here as a human.
Data.
Search by name.
Yeah.
Name and description.
These are OK.
I mean, they're a little bit overlapping in terms of what they are saying here.
So this isn't how I would write them myself.
This is where I would say that I can search by different fields.
The ones that are named earlier were the data splits.
So rules, probably nothing on rules specific here.
Can you talk about feature splitting is a question in the chat.
Anything you want to add on that for a quick second?
Possibly.
I'm not sure what you mean by feature.
A lot of times people will use the term feature to mean a story that's big enough to be released.
So high level story term Hunter used earlier.
And this is not This tool works on stories of whatever size you want to put it.
So you can give it a big story, small story.
And we've talked a little bit about, can it with enough context be not just a splitter, but kind of a backlog idea generator?
And I'm not sure I see any problems with it going in that direction.
You'd want give a bit more context on the higher level stories.
But if you mean feature as high level story, it should work the same as it does on smaller ones.
And some of the ones we're seeing here are pretty small.
So.
A couple of things.
You do a demo how to input the story.
I'll show that in a second.
It's not that interesting.
Just because it's quite simple, but yes, I will show it.
Even loading them up just because they take a couple seconds to run and don't want to delay folks.
There's a note in here about mountain goats AI is great.
I think maybe what that person Ron, maybe you're talking about our goat bot tool.
So for those that don't know, we do have another free tool that's been out and available for a while called goatbot, which is basically like an agile coach.
It's got everything Mike's ever written in all of his courses and books and whatnot, and weekly tips and blog posts.
And it uses that to try to sort of answer general agile questions.
That's available to you to sign up for.
Now too, if you want to, that's on the mountain goat site.
I think it's just, If you can go to goatbot.mountaingoatsoftware.com, you'll find it.
It's also in our regular website navigation.
You should be able to find there without too much trouble.
If that interests you, this is a more specialized tool for doing story splitting.
So let's see.
All right, I'm going to skip forward here and we'll pick another one that we haven't done here just to get it in.
And oops, sorry, wrong thing.
I'll open this up.
So the actual interface here is quite simple.
There are some examples.
there's a space for you to enter your story.
And then there is a place for enter some acceptance criteria.
That's really all that there to it, at least at this point.
Again, prototype tool that we're trying to get feedback on.
Some ideas about ways that you could add some additional pieces of information you might be able to put in to give you better responses.
But for now, it's relatively free form.
You can include information in here, like saying, oh, this team I'm working with isn't very familiar with this technology,
and that might influence more spikes, for instance.
So there are some levers you can pull to try to influence your results.
Let's just put this one in and have it go.
You see here it starts doing things.
CS in the chat, yes, the spikes, paths, interfaces, data rules is the grouping of five rules that I came up with for ways to split stories.
And so what we've done is we split those and we trained it on those, and then for convenience in reading the results, hunters displaying them like,
hey, here are the things we found by splitting or bypass.
Here are the things we found by interface, by data, my rules.
So that's the grouping that we spit out based on the training that has been done on those five specific splitting techniques,
which is why they can be overlapping.
That's why in some I was saying, oh, I keep the path ones here, but get rid of the data ones, right?
They're fully expect these to overlap.
So here was the result.
The story was, as a user of an application, I want to log into the application so that I can start using the applications.
So what do you think about this, Mike?
Yeah, the paths I think are great.
I mean, this is exactly the way I would split this for a path.
Social media, I can do it through email and password.
Interface, again, kind of what we'd expect there with just kind the different user interfaces.
Can you scroll down to?
Yeah.
Scroll down the data here.
Log in with my username and password.
Again, there's a good example of an overlap.
Same thing, right?
What I would...
What I would expect a person to do for the original story, and Hunter, maybe scroll back up the originally story.
What would I expect the person do with the origin story here, if I saw this, would probably be to think of all the things about logging in,
which would be, I'm locked out after three failed attempts.
Actually, let me change that.
I am locked after N failed attempt.
I'm required to enter a strong password.
I can recover my password somehow.
To me, when I see a login story, I think about all those sub elements, and I've missed it with those here, but we didn't really tell it to think with the
story that we gave it.
I don't know if we went into the acceptance criteria and, you know, reran it and just said, include all the normal things for a password story.
Or if he said include things like password reminders, strong passwords and things of that, would we get all of the splits that I would anticipate?
Let's see, a couple notes from chat questions.
Do you have to tell it that's acceptance criteria since it's just an empty text box?
You don't.
It assumes that it is.
This is kind of part of the prototype user interface.
You know, I can imagine being much more specific and structured in the fields that we present versus that sort of open, free-flowing text field.
That's kind part what we're experimenting with to see what is easiest for folks, the best way to provide this stuff.
Right now you don't have to say it and you can kind of just write it, and it will.
It's supposed to just figure it out.
So that's the way it's currently set up.
Let's see, will this work on copilot?
I saw this question a few times, so this is its own tool running on our own infrastructure using using those models we were talking about earlier.
As far as will it work?
On copilots, I guess I would like to turn that question around.
People are using AI systems inside their companies that they would We'd like to know more about that.
I'm just curious what you're using and how you use it, and maybe you have rules about what your allowed to use and you are not allowed use external tools.
As we decide what we're going to build and what were going focus on, that kind of info is actually really useful to us.
So I don't know if you want to mention it in the chat, just anything that you wanna share.
That kind stuff is interesting for us, so.
And just one other note on spikes.
Yeah, I've noticed too that we haven't seen too many.
The sort of default instruction is that the team is competent in the tech that were asking them to use here.
So unless specified otherwise, it assumes that they're pretty good on the text side.
And I think that influences its decision to spike or not spike in a lot of cases.
We'll see if we get any of these others that do have a spike component.
Yeah.
I know it's a lot easier to copy and paste ones in and you didn't want me to give you ones to type, but this will be short.
Can you type one in for me and we'll see what happens?
Yeah, so I'm curious on this one.
So what if we put in as agile team member, I want an artificial intelligence to split stories for using the spider and invest approaches.
And we could add it so that if we wanted, but that's probably enough.
So we couldn't put spider in invest in acceptance criteria.
Would that be better?
Do you think?
I'm trying to get it to generate a fight for us.
Yeah.
Well, why don't we say the team is very new to this.
See what that does.
All right, we'll give it a second to do its thing.
We haven't posted the tool yet, so we will post the link before we're done, I promise.
So I'm happy if folks are excited to give a try.
If you haven' missed it, it's not a secret.
We're for the most part just doing this as a fun project for now.
And we fully intended to post the link unless everybody looked at it today and said, Oh, that sucks.
Please don't post that way.
So exactly.
We were written.
I mean, maybe not say anything, but it seems like people are interested enough to give it a go.
OK, so we got a spike on we want to write in AI for splitting.
Investigate and prototype an AI-based solution that uses Spider and invest to automatically split.
Sorry, it doesn't tell us anything there.
The spike should include a review of existing techniques and evaluation integration challenges and identification of core architecture requirements to
reduce uncertainty before.
I think that's good.
I would be very happy if I was part of a sprint planning meeting and somebody said, hey, let's do an AI to do this.
And we clarified it with that three sentences or whatever that is there.
I think that would very helpful if were the person to go off and work on that.
Especially given the team's limited experience.
Warrant to the single spike.
Yeah, so I that's a very good answer there for Spike.
So hopefully that helps.
Path.
I can input a story and choose the Spider method.
Automatically generate, make sure it meets invest.
It knew to choose Spider as an approach and that invest is something to meet.
You meet the invest criteria, so it knew that.
Here's an example, one that I wouldn't necessarily have thought of the third path one there, I can access onboarding guidance about those approaches,
right.
Would I not have thought about that?
I did not think about.
Now, we haven't really polished our tool, but I don't think in the tool we have a button like Explain Spider to me or Explain Invest to Me.
And would I have that when we get to the point of doing a fancy UI?
Maybe, probably would have eventually.
But I wouldn't have though about it early on, and it would've been a surprise, then the dev team would had to scramble to put it in when I didn't.
So that's an example of this helping me, I think.
Interface is lazy.
Here's your three platforms.
Well, it is weird because they're very wordy versions of the same thing, right?
So I don't know why I got so word on do it in iOS and Android.
Minimal data set.
Yeah, doesn't really tell us.
Oh, I guess minimal data said mean we don' train it on a whole lot so we can start it.
That's what it means.
So i think that's pretty good.
I think thats pretty go.t I was surprised by that.
I would have done that, but I wouldn't have thought about it as a separate story of like, hey, train it on 100 stories to start and 1,000 later or something.
Couple notes here from folks.
Oh, some interesting stuff on copilot.
If I think about copilots, since Microsoft's used a brand for so many different things, it's hard to know exactly what people are talking about.
Yeah, I mean I use GitHub Copilot all the time myself.
And I know that they have tools built into Office and I guess there's a DevOps tool that I didn't know about, so good to get to that that's something that
people were interested in for sure.
So, yeah, I think let's continue.
We're going to try to get through some more stories here.
I don't think we're gonna get to them all because we've got quite a few, but fortunately, if we don' get yours, you can try it for yourself.
Hunter, we did have a question about, can this work in conversational mode?
And we do let them go back and revise the story, of course, This is, no, there's not in the current tool, so we have a Goatbot tool which is a sort of
more traditional conversational agile coach chatbot type tool.
It also does sort put limits on the length of the conversation to just try to keep things concise.
This not a back and forth style thing.
I've thought about adding some additional features like maybe one to say, give me more of, the give you more path splits or now try again.
Ways to kind of either generate more without regenerating the same stories right so it knows has the context of what it's doing.
So none of that stuff exists yet again, this is stuff that we're actively playing with and trying to get a sense of.
what people want, how it works, and how they might use it.
That's interesting feedback.
We appreciate that.
It is separate from GoPot at the moment.
Maybe there's a feature where they end up as different modes in the same tool.
Again, we're not exactly sure where this is all leading.
Or we are experimenting.
As a database admin, I want to be able to query the database for objects with the failed status so they can quickly identify objects that need to resolved.
Oh, so this is a good example.
We had a request for something that was much more technical and not one that would be about my mom wanting something, right?
So this a Good example of that.
No spike, probably agree with that as a database admin, I want to execute a preset query.
Yeah.
I guess, OK, we've saved a query to find those if this is something that sounds like somebody might be doing frequently.
I'm not sure that really needs to be a thing as much as the system lets us save queries.
As a database admin, I want to refine my query parameters.
That's probably a good one.
Right?
I don't want find things that failed in the last week versus the seven hours or something.
that's Probably a decent one And then interesting here that it came up with better, not better but different interface suggestions.
So I quite like that.
I mean, I was expecting to see some silly iOS Android nonsense here.
so I think this is quite nice on the interface that put those in.
Data, I want to query the data set for objects with failed status.
Want to enhance it.
These are good.
I think these are pretty good, so I did well on this technical one.
So yes, that's why Spikes interfaces data.
Spider is a technique that Mike came up with that is in our Better User Stories course, something that we teach to try to help people do story splitting.
That's what's reflected here.
Let's apply those for chosen.
Yeah, what I did is, as I mentioned, I had 1,100 and something stories.
I have this huge office because it was really cheap.
It was underneath a dentist, by the way, never rent office space under a dentists office.
But I thought it would be good.
And it had a big wall, but that was about it.
So I taped all these stories of all over the wall and I started grouping them.
Had like eight groupings.
Was trying to group them into the smallest set I could at one point.
Then I was trying kind of form an acronym around them At one point, the acronym was Paris.
And you can see some overlap between the letters in Paris and Spider.
I don't remember what the A would have been anymore.
But anyway, I ended up with Spider as the set that when I tested it then against another 100 or two stories, that I was always able to find a split.
So there's certainly other ways to think about splits.
Any way you think of splits are not necessarily mutually exclusive.
You might come up a A split from this approach or this or approach, but you came up the same split, and we're seeing that here.
It's just a matter of having a thinking tool.
Okay, think about the paths through a story.
It came up with some splits.
Think about some rules the story has to support.
And if you need 13 ways to think, great.
My brain's not big enough to remember 13 things, so I wanted to have a small set.
That was the five.
With those five, I found that I can always find a way to split something.
Here we've got another one.
As an Alexa user, I want to be able to create a schedule that has repeatable events and unique events with multiple types of reminders so that I can manage
my life better.
I always have to remember with stories about this particular Amazon device that I can't say her name out loud or it starts to activate in some cases.
I have a friend who's a popular Apple podcaster and he, instead of saying, Hey S I R I, he says, hey dingus out on his podcast.
So because he doesn't want to wake up everybody's home pods and Apple watches and whatnot.
Yep, we have an example about her in our CSM class or something like that, and we've had it go off before.
And the example is silly.
It's about having it work with your dog.
Your dog can bark and activate the device, so your dogs wants to order steak, I worry about ordering steak accidentally.
So a spike here for creating a schedule.
Investigate, document the technical feasibility for her scheduling.
Yeah, that's probably a good spike.
Research what's available there.
Paths, I want to create repeatable schedule events.
want to create unique one-time events, I like these.
I liked the path splits there.
Interfaces looks the same as path in this case, so I don't see any.
Yeah, they're a little bit more similar than I would like.
That, not to get too in the weeds, that's probably an implementation thing that can be fixed with some tweaking, but I've seen that before.
Yeah.
If we can make those better, we always want to do that.
But again, my point with the five approaches here, or more if we want add more, is that it's five ways to come up with a split.
And if three things all lead you to the same split, fine.
It doesn't really matter.
The goal is you thought of a way to split the story.
How you got there doesn' really matters.
So let's look at the data ones.
Create a schedule that's both repeatable and unique.
with a basic default reminder.
That's a good one that would be interesting to give it as a sub-split.
Take that story, that first data story and split it further because I would hope it would do things like I can do repeatable ones,
I could do unique ones.
I consider basic reminder or I considered advanced reminders which might be like, tell me a day ahead and tell an hour ahead or something like that and
let me configure those.
I want to enhance my skills of events by a multiple type.
Yeah, multiple types of reminders.
So there's the one I was just thinking about.
Those are good.
I would say this is one of the better splits that we've had.
Hunter, any idea why that's one the other ones?
Not really.
That's kind of thing.
It's hard to say without digging into the details.
To analyze some of explanations might be to see, was it the amount of context that was provided?
I don't know.
I have a hunch on that one.
But that is something where we might want to dig in.
And that would be the kind of thing where if you've got a split like this, and you're like, this is great, you write, well,
we've gotta if we haven't shown it, because it's the bottom of the screen, got this little feedback form down here.
When you when you try this for yourself, just to give us a sense of, you know, generally how you thought it was.
And really, if you can, I mean, it takes a little bit more time.
But if You can give a comment, that's really helpful in terms of figuring out tuning for the future, because whether it's good or bad is interesting.
but we're saying, hey, i really thought this needed to be better is is really useful.
So Yeah, and that came about from me giving feedback on a lot of the initial ones, where it was like, I don't know what to say,
because the paths and the rules might have been amazing, but the interface one sucked or something, right?
What score do I give it?
So made it hard.
And Harry, you commented about adding business rules.
Yeah.
The more we put in those kind of acceptance criteria, the more rules there that would help it spread stories.
Definitely.
And I realized you guys didn't know in advance kind of what we're going to be asking you to provide.
You may have provided them in your examples if you had known that.
But yeah, I have found that if do provide some of that additional information that you will get some richer responses.
Well, we are limited with the chat length, too.
Hunter, We did have a question here, which has to do with what I was commenting a moment ago.
Can we feed one of the splits in as a story?
Would it be possible to send in the split that we had a moments ago?
There was one that I liked.
Which one?
I think it's one these, right?
Yeah, it was the first data one, but let me see if that's the best one.
Yeah, just try that one.
All right.
I'll just copy and paste that in.
Forgive my manic scrolling.
Whoops.
Got a bunch.
So that's going now.
Give it a moment.
We'll see what happens.
Oh, yeah, go ahead.
I would expect to see stories, a story about repeatable story, about unique and story.
About a basic reminder and then a.
Story about advanced ways to set reminders.
So, so we got the same spike.
Okay.
In our world, presume we've already done that spike, So that's not needed.
But it wouldn't know that.
I mean, so I'm just sorry to interrupt you for a minute.
A feature, an actual dedicated sort of resplit feature probably would feed that existing context in to prevent it from just repeating itself.
Being like, hey, this is what you said last time.
Don't do that again.
That would be nice.
Yeah, that would be nice instead of just doing what I've been doing, which is what you did.
Just, you know, retype it in.
But that will get a little bit of that conversational aspect, right?
So there we go with the story about a one-time event and repeatable event.
And both of those have a basic reminder.
I think that's fine.
So this is basically the same.
Yeah.
I mean, it's using the voice as the interface, but that's, you know, that.
That's OK.
Can you see the actual story again back at the top?
Repeatable and unique.
OK, so I guess we've told it to only give us a basic reminder system, right?
And so that's where I was wrong in my assumption, where said I'd expected to see basic reminders just like, you know, tell me x minutes ahead of time.
But I want to also expect to say, let me set it be a minute ahead, an hour ahead a day ahead.
Things like that.
I just scheduled yesterday a doctor appointment.
The doctor was like, come back in a year.
Let's schedule that now.
Like, OK.
And so I put in like a reminder of my calendar one month ahead of time.
Just make sure I still want to go see that doctor.
So I would have expected the fancier scheduling here, but we kind of specifically told it not to in the story.
Tracy has a question.
I don't have an answer, but I have another question, so it's a.
Question about details on data processing data use so that they could run it side by side with their other tools at work.
Um, I.
Don't.
Have anything like that right now.
Uh, understand that that's the concern for some folks.
as many people to be able to use this in their work as they can.
It's a work tool.
So that sounds like something we could consider, but it'd be great to get some kind of general feedback on what are the powers that be looking for in that
regard in order to acceptable.
Yeah, we've thought about what type of additional data security and stuff like that we have on there.
And if we ever do anything with this where it's like we're really wanting you to use it big time, you know, will definitely look into those things for now.
It's just put in a couple of your stories out of context where there's nothing that anybody's ever going to steal.
I mean, just try it out and let us know if it seems interesting.
That's kind of all we want to see if this kind passion project is worth is worth pursuing for people.
Let's see, we've got a few more minutes.
We have to make sure we remember to share the link with you guys.
So well, yeah, I'll do a couple more.
Will share.
The link and then we'll let you go and you can give us feedback in the system itself, so close that one.
I think this was the original one OK.
Let see as an IT administrator.
Want to migrate our base back based services to Azure DevOps to take advantage of scrap.
Excuse me, cloud scalability and reliability.
OK, let's see.
Hunter, I might want some of your comments on this one as well.
You're doing something similar right now with some tech migration.
I like the spike.
Oh, did see a question there.
Does it ever generate more than one spike?
In earlier testing that I did, it did.
See, am pretty sure I saw more then one spiked.
Am almost positive I see more that one spikes, but not very often.
I have answered quite a bit and probably used it more than anybody else.
I've seen a few times where it's done more that one spike.
It seems to often not though.
And I don't know if that's a question of some of the tuning, some the base level tuning that is in there as far as, you know,
the team's confidence is one and there's few other things that are assumptions that were made.
So that may be part of it.
I think it's more that if I just think about the times with real teams, we've rarely had two spikes going on on the same story.
If we did, would kind of roll it into one spike.
Go do the research on such and such.
The times where I can think of doing concurrent spikes on a story were more about tech feasibility and desirability of the future.
Let's see if people really want this thing.
So that's where I go with that one.
Down for path.
Let's see here.
IT administrator.
Oops, my screen is refreshing.
I want a guided migration wizard.
That's nice.
Automated build and deployment pipelines.
I don't know that these, oh, I'm sorry, looking at interface.
I'll stick with interface for a second.
These would be the ones that would survive human review, but they seem like they would give a person good ideas about writing the right story for these.
They would probably give us a jump start on those.
The path ones.
Migrate back-based services source code to DevOps.
after migrating the source code.
The second one's interesting where it's a little bit of a job story where as a person after something, so it has got a win in there.
It's not just as an IT administrator, I want to set up automated build pipelines, it is knowing that it had to do it after my migratory source.
Same with the third one.
What do you think about those, Hunter?
Do you have any thoughts?
They seem reasonable, pretty good.
I mean, maybe a couple of them are better than others, but they seem like reasonable starting points for some of this stuff.
Me not being super familiar with all the stuff that's in the Azure DevOps bucket, they seemed plausible.
Yeah.
Okay.
So I think those are good, and those were pretty.
Better than I thought it was going to do on that story.
Okay, let's see.
As a driver, I want to be able to easily access my media or streaming accounts in my vehicle so that I am not bored on long journeys.
That's a good story.
Yeah, that spike is interesting because who are we, right?
This one probably, and I haven't looked, anything below here yet.
But Are we trying to come up with mountain goat play?
We're going to put it in cars instead of car play in Google.
So are we out there investigating the whole world and figuring out what's out?
Are building on top of what there?
So I think the spike is interesting.
And that would depend on who this company is.
As a driver, I want to log into my streaming account.
Switch between multiple streaming accounts directly.
Yep.
In the interface, I'm going to hope there's one down here about voice.
We'll see that in a second.
Driver I want to control media playback through intuitive dashboard controls.
Yeah, these are good.
Yes, and interface it did pick up that there should be a voice command as well as the in-screen touch screen.
So I've been happy with some of these that As much as I was kind of down on like it was always an iOS and an Android story and a browser story,
it has on the ones that that's not relevant.
It has picked that up.
Right.
So I think that.
That's been good.
Driver, I want to connect my primary music streaming account, for example, Spotify, my vehicle, the simplified.
One of your additional media accounts.
Yeah, I like that.
I might want to have a story in here.
This might be more about rules where I wanted to default to using title.
that's my preferred one.
Most of my playlists are on title, let's say, and so default to title.
That's a little bit implied here by saying there is a primary music account, but I might want to explicitly say that I can do that on my home.
I use a Roon music player at home and I could set that up there, look for here first.
So I'm not going to have a story like that.
It did, I think you did a good job here.
So why don't we give people this URL so they can go sign up?
So I'm going to put it in chat right there.
You should see it.
It's on our website there So you can go and get access to this tool, it's free.
And in addition to that, let's just do a quick reminder that we've got another webinar next week that you could sign up for right now.
It's all about user stories, not just splitting.
There's a whole bunch of different content.
That's about an hour long as well, plus a Q&A.
So if you wanna go in depth on user story stuff with us next we'll be there.
Yeah.
I want to thank everybody for being here.
Thank you to Hunter for putting all this together so we could do this.
And look forward to your feedback on this Love It, Hate Us.
Love it, hate it.
Hate us, I don't want you hate us.
But love it or hate the tool, let us know.
Curious on what you think about that or things that we should be training on to make it better.
So greatly appreciate your time.
Thanks, everybody.
Yeah, thank you guys for the feedback.
I got some really good feedback in here and I look forward to sort of going back through the chat and also, you know, feedback on the tool.
And, please be honest, like I said, it's a prototype.
So we want to see what you think.
All right.
Thanks all.
Bye.
Let's do this.
